Text Classification
MLX
Safetensors
English
qwen3_5
qwen3.5
decision
unofficial-port
experimental
4bit
4-bit precision
Instructions to use cowWhySo/jeeves-mlx with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- MLX
How to use cowWhySo/jeeves-mlx with MLX:
# Download the model from the Hub pip install huggingface_hub[hf_xet] huggingface-cli download --local-dir jeeves-mlx cowWhySo/jeeves-mlx
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
- Atomic Chat
Download validation.json from cowWhySo/jeeves-mlx: direct link, hf CLI and curl.
- Browser
- Download file 2.25 kB
-
https://huggingface.co/cowWhySo/jeeves-mlx/resolve/main/validation.json
- Command line
-
hf download hf://cowWhySo/jeeves-mlx/validation.json
-
curl -L -o validation.json https://huggingface.co/cowWhySo/jeeves-mlx/resolve/main/validation.json
2.25 kB
| { | |
| "variant": "jeeves-mlx-4bit", | |
| "identity": "c4d0ea30e584dbf0d1e672c59c2e00036cb6a1a7025833ef49c01c63432b18fa", | |
| "conversion": "weights_converted_and_structurally_checked", | |
| "format_head_checks": { | |
| "status": "passed", | |
| "format_cases": 4, | |
| "original_head_dtypes": { | |
| "q.weight": "torch.float32", | |
| "q.bias": "torch.float32", | |
| "k.weight": "torch.float32", | |
| "k.bias": "torch.float32" | |
| }, | |
| "exported_head_dtype": "float32", | |
| "temperature": 1.8589280843734741, | |
| "max_head_probability_error": 0.0, | |
| "scope": "Original encoder token/readout parity, safe head serialization, and original PyTorch head vs MLX head on synthetic hidden states. Not backbone parity." | |
| }, | |
| "cpu_smoke": { | |
| "status": "passed", | |
| "identity": "c4d0ea30e584dbf0d1e672c59c2e00036cb6a1a7025833ef49c01c63432b18fa", | |
| "device": "cpu", | |
| "cases": [ | |
| { | |
| "id": "case0", | |
| "probabilities": [ | |
| 0.016998127102851868, | |
| 0.9830018877983093 | |
| ], | |
| "elapsed_seconds": 4345.674271831 | |
| } | |
| ], | |
| "fixture_sha256": "2887c941fd5a497ae51343a9e2382ea1d198ab5eb140f4e8c8fd37c54be4ebf0", | |
| "scope": "Small no-thinking execution smoke test; not a benchmark." | |
| }, | |
| "bf16_backbone_parity": { | |
| "status": "small_fixture_check_passed", | |
| "cases": [ | |
| { | |
| "id": "case0", | |
| "maximum_probability_delta": 0.00010848045349121094, | |
| "minimum_hidden_cosine": 0.9999632468948669, | |
| "top_option_agreement": true, | |
| "within_diagnostic_tolerances": true | |
| } | |
| ], | |
| "probability_atol": 0.02, | |
| "minimum_cosine": 0.995, | |
| "scope": "Heuristic diagnostic tolerances, not bitwise equivalence or general accuracy evidence." | |
| }, | |
| "apple_metal_validation": "not_run", | |
| "task_accuracy_calibration_benchmark": "not_run", | |
| "speculative_drafters": "not_ported", | |
| "greedy_generation_validation": "not_run", | |
| "quantization_vs_mlx_bf16": { | |
| "cases": [ | |
| { | |
| "id": "case0", | |
| "maximum_probability_delta": 0.016297996044158936, | |
| "top_option_agreement": true | |
| } | |
| ], | |
| "status": "measured_not_quality_certified" | |
| }, | |
| "overall_status": "converted_and_small_cpu_diagnostics_passed_not_benchmarked", | |
| "diagnostic_failures": [] | |
| } | |